Selected Work
Projects
Production AI systems, LLM pipelines, computer vision, real-time backends, and IoT platforms — shipped end-to-end across healthcare, hospitality, fintech, and consumer tech.
Healthcare · IoT · SaaS
upLYFT
Physical Rehabilitation Platform
Two-sided physiotherapy SaaS on Azure. BLE wearables stream 200Hz+ sensor data to a cloud ML pipeline computing joint angles, gait symmetry, power, and load in real time. Clinician, athlete, and admin dashboards with sub-100ms latency supporting 500+ concurrent sessions.
FastAPIFlutterBLE 5.0AzureView case studyLLM · Real-Time · Hospitality
QuickComm
AI Hospitality Communication Engine
Real-time AI communication backbone replacing walkie-talkies in hotels. PCM audio ingested from radios, transcribed at 94%+ accuracy via Whisper/DeepGram, classified by Gemini LLM at ~88% intent precision, and routed to staff teams instantly. Monolith-to-microservices migration delivered 3x throughput at ~$3/month per property.
FastAPIGemini APIDeepgramWhisperView case studyIoT · Wearable · EdTech
AI Pendant
Smart Wearable for Kids
IoT pendant wearable for children with 160+ API endpoints. Device pairing, Pomodoro timers, missions, mood tracking, sleep monitoring, baby-cry detection, rewards and badges, parental dashboard, educational content, and anti-loss alerts. Gemini-powered emotional insights and activity analytics.
FastAPIPostgreSQLGemini APIFirebaseView case studyFinTech · IoT · Mobile
The Giving Cube
IoT Donation Ecosystem
Complete IoT charity platform connecting a smart donation box, mobile app, and web dashboard. Physical coin/note donations auto-detected by the smart cube, processed via Stripe, and synced to charity dashboards in real time. BLE/WiFi dual-mode with 99.9% transaction reliability.
FastAPIStripePostgreSQLBLEView case studyAI Vision · Generative · Product
Interior Design AI
AI-Powered Room Styling
Photorealistic AI pipeline for furniture placement in room photos. Three-stage pipeline: item gallery generation → AI mood board with Gemini → composite scene rendering using depth-aware perspective scaling. Painter's algorithm compositing places furniture at accurate scale with shadow blending.
PythonPillowOpenCVReplicate APIView case studyComputer Vision · AI · Security
CCTV Anomaly AI
Real-Time Surveillance Intelligence
Multi-phase AI surveillance system evolved from prototype to production MVP. YOLOv8 detects crowd surges, loitering, intrusion, and fights across 8+ simultaneous camera feeds. WebSocket streams push annotated alerts to operator dashboards with configurable per-zone thresholds.
YOLOv8OpenCVFastAPIPythonView case studyMobile · GPS · Sports Tech
Apollo Golf GPS
Smart Caddie Companion App
GPS-based golf companion app replacing rangefinders for 1,000+ active users. Backend handles ball tracking, scorecard management, and shot analytics. Optimised PostgreSQL schema delivers course data in under 80ms. Integrated real-time GPS with ±2m accuracy for precise on-course navigation.
FlutterFastAPIGoogle Maps APIPostgreSQLView case studyEnterprise · Digital Twin · SaaS
MRO Digital Twin
Automotive Workshop Management
Full-stack digital twin for automotive MRO workshops. Seven role-based dashboards (manager, reception, technician, store, QC, QA, supervisor) covering the complete vehicle lifecycle — intake → work order → stage checklists → spare GRN → QC → delivery. 34+ database tables, barcode/QR integration, real-time KPI dashboards.
FastAPIReactPostgreSQLSQLAlchemyView case studyAI · LLM · Research
PaperIntel
RAG Research Intelligence
End-to-end Retrieval-Augmented Generation system for research papers. PDF ingestion, chunking, BGE embedding, ChromaDB vector store, cross-encoder reranking, and citation-aware generation. Supports similarity, MMR, and hybrid BM25+vector retrieval with query expansion and multi-hop decomposition.
FastAPIChromaDBSentence TransformersPyMuPDFView case studyLLM · Real-Time · Audio
Real-Time Translation
Multilingual Audio AI Platform
Real-time translation backend capturing live PCM audio, transcribing and translating into 10+ languages in under 800ms end-to-end. Achieved 96%+ transcription accuracy using Whisper with custom language model fine-tuning for domain-specific vocabulary.
FastAPIWhisperPythonWebSocketsView case studyIoT · AI · Sports Tech
Smart Boxing Gloves
AI Combat Sports Analytics
Real-time AI system embedded in boxing gloves tracking speed, force, calorie burn, and motion patterns. Hardware sensors stream data to cloud at 200Hz+; ML models analyse punch mechanics and training load. Data-to-insight latency reduced to under 50ms with hardware-cloud integration.
PythonFastAPIWebSocketsESP32View case studyLLM · Voice · AI Assistant
BadarAI Assistant
Conversational Voice AI
Full-duplex voice AI assistant with streaming STT, LLM reasoning, and TTS synthesis. Deepgram streams speech-to-text in real time; Gemini handles multi-turn dialogue with persistent memory; ElevenLabs renders natural-voice responses under 800ms. Supports persona customisation and live tool-calling for real-time data fetching.
FastAPIGemini APIDeepgramElevenLabsView case studyNLP · AI · Fraud Detection
Fake Ad Detection
NLP Fraud Classification System
NLP fraud identification system classifying 50,000+ online advertisements. Ensemble approach combining LSTM, SVM, Doc2Vec, and TF-IDF features achieves 94% classification accuracy with 40% faster inference compared to baseline models. Deployed as a FastAPI microservice with real-time scoring.
PythonLSTMSVMDoc2VecView case study